Noise robust exemplar matching with alpha-beta divergence.

Emre Yilmaz,Jort F. Gemmeke,Hugo Van hamme

Speech Communication（2016）

引用 5|浏览86

暂无评分

摘要

•Noisy speech is modeled as a linear combination of multiple-length exemplars.•The proposed recognizer uses the alpha–beta divergence with two parameters.•The parameters are automatically adjusted to obtain better separation.•The adaptive noise modeling approach is effective on the genuine room noise.•Proposed recognizer provides improved noise robustness on AURORA-2 and CHIME-2 data.

查看译文

关键词

Automatic speech recognition,Noise robustness,Exemplar matching,Alpha–beta divergence,Reconstruction error

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要